September 2004 - Boost-users

boost.mem_fn with _com_ptr_t
by Andy Tompkins 17 Sep '04

17 Sep '04

Hi, I am trying to use com smart pointers with boost::mem_fn I defined the following so that mem_fn would know about _com_ptr_t s: template <typename Interface> inline typename _com_ptr_t<Interface>::Interface* get_pointer(_com_ptr_t<Interface> const& p) { return p.GetInterfacePtr(); } but when I use the function object returned from boost::bind, I get a null pointer in my _com_ptr_t ADODB::_RecordsetPtr recordSet(__uuidof(ADODB::Recordset)); ... boost::bind(&ADODB::_Recordset::Close, recordSet)(); The problem seems to be in mem_fn_template.hpp The function class BOOST_MEM_FN_NAME(mf0)::operator()(U &u) const calls BOOST_MEM_FN_RETURN call(u, &u); and _com_ptr_t::operator&() calls _Release() and sets m_pInterface = NULL; Is there any way to get this to work correctly? Thanks in advance. Andy.

1 0

Re: [Boost-users] sort implementation based on swap
by John Meinel 17 Sep '04

17 Sep '04

Howard wrote: > > On Sep 15, 2004, at 5:28 PM, Russell Mok wrote: > >> I know this is kinda off topic. >> >> I am trying to sort some objects which are expensive to copy. Does >> anyone know a sort implementation that uses neither copier nor >> assignment? The STLport version I have is using both. Worse, some >> routines are passing arguments as T instead of const T &. I did >> search boost and google but I couldn't find what I want. > > > The Metrowerks implementation of std::sort uses swap (unqualified) > exclusively. > > -Howard > > Another idea which I think works out for simplicity is to just sort an array of pointers, and then once all items are in order, you change the ordering on disk. I think this would be your biggest performance gains, as most sorting has a lot of intermediate steps. You can do all those fast with the pointers, and then do the slow sort of the files. John =:->

1 0

Newbie help in compiling boost
by Murali Krishnan Ganapathy 17 Sep '04

17 Sep '04

Platform: Windows XP Compiler: VC7 (Visual Studio .NET) Shell: Cmd.exe BJAM: Prebuilt version 3.1.10-1-ntx86 from sourceforge Boost Version: 1.31.0 On the prompt: * set PYTHON_ROOT=C:\Python23 * set PYTHON_VERSION=2.3 * C:\Program Files\...\VC7\bin\VCVARS32.BAT (execute this) * bjam "-sTOOLS=vc7" install I get a whole lot of lines of the form ...skipped XXX for lack of YYY... and finally get ...failed updating 2 targets ...skipped 3077 targets Any pointers on what I am doing wrong. On the prompt I am able to run "cl, link, nmake" without any problems (i.e. path is set correctly by vcvars32.bat). - Murali

3 4

RE: [Boost-users] Boost problems
by Puverle, Tomas (IT) 17 Sep '04

17 Sep '04

> > We have found some problems with the following files: > > > > boost_1_31_0/libs/regex/src/cpp_regex_traits.cpp > > boost_1_31_0/libs/test/src/test_tools.cpp > > > > They use definitions from <ios> but never actually include it > > - they are > > relying on the fact that in some STLs the headers are pulled in by > > other header files. Unfortunately, under our STL (RW STL > 6.0) this is > > not the case and hence the files fail to compile. Add the > following > > include line to both to fix the problem: > > > > #include <ios> > > Ok. > > Gennadiy. Gennadiy, have you any news about the second problem also related to regex, as described in my original email? Thanks, Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

1 0

[boost-users][MPL] Where did eq, gt, lt, et al. go?
by Cromwell Enage 16 Sep '04

16 Sep '04

These metafunctions used to be defined in the same headers as equal_to, greater, less, etc., respectively. What functionality should I use to replace them for my sequence algorithms? Cromwell Enage __________________________________ Do you Yahoo!? Yahoo! Mail - Helps protect you from nasty viruses. http://promotions.yahoo.com/new_mail

2 1

boost.graph, subgraph questioon
by Jeffrey Holle 16 Sep '04

16 Sep '04

I'm using boost 1.31.0. I've experienced and observed that other fokes found this problem with subgraphs and internal properties. Derived subgraphs only get internal properties produced by default construct. What I'd like to know is if adding an edge to a derived subgraph via: add_edge(vertex_descriptor u_local, vertex_descriptor v_local, const EdgeProperty& p, subgraph& g) will get the property sent to the ancestor graphs? Are there any other "bugs" associated with adding edges/vertices to a derived subgraph?

1 0

It seems that boost.random doesn't work under vc6.0?
by huangliang 16 Sep '04

16 Sep '04

When I tried to compile the random_demo.cpp using vc6.0 and stlport5.0, there were many errors and warnings. Does boost.random correctly work under vc6.0? I use boost1.31.0.

1 0

RE: [Boost-users] Boost problems
by Rozental, Gennadiy 16 Sep '04

16 Sep '04

> -----Original Message----- > From: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Tom Puverle > Sent: Thursday, September 16, 2004 4:53 PM > To: boost-users(a)lists.boost.org > Subject: [Boost-users] Boost problems > > > Hey boosters, > > We have found some problems with the following files: > > boost_1_31_0/libs/regex/src/cpp_regex_traits.cpp > boost_1_31_0/libs/test/src/test_tools.cpp > > They use definitions from <ios> but never actually include it > - they are > relying on the fact that in some STLs the headers are pulled > in by other > header files. Unfortunately, under our STL (RW STL 6.0) this > is not the > case and hence the files fail to compile. Add the following > include line to > both to fix the problem: > > #include <ios> Ok. Gennadiy.

1 0

sort implementation based on swap
by Russell Mok 16 Sep '04

16 Sep '04

I know this is kinda off topic. I am trying to sort some objects which are expensive to copy. Does anyone know a sort implementation that uses neither copier nor assignment? The STLport version I have is using both. Worse, some routines are passing arguments as T instead of const T &. I did search boost and google but I couldn't find what I want. Thanks in advance. Russell

3 3

Edmund Karp sample
by Daniele Carlucci 16 Sep '04

16 Sep '04

Hi all, how can I print only the out edges from a node in this sample? I only print 0->1 1->0 2 3 2->1 3->1 there is a way to print this output: 0->1 1->2 1->3 Thanks Daniele Carlucci

1 0

thread (read_writ_mutex and recursive)
by Gordon Smith 16 Sep '04

16 Sep '04

Are there any plans for a recursive version of the read_write_mutex? Gordon.

2 1

Building boost graph library for Visual C++ 7.1
by Brian Parker 16 Sep '04

16 Sep '04

Hi, I have performed a fresh install of Boost 1.31.0 under Windows (Visual C++ 7.1) and everything built and installed OK except that the Boost Graph Library libraries needed for GraphViz interface (libbglvis.lib) were not built or installed. Does anyone have any suggestions for building this library? I am not familiar with the jam file format. Thanks, Brian Parker

2 1

program_options (Use "/" instead of "-")
by Gordon Smith 16 Sep '04

16 Sep '04

Does program_options support use of "/" instead of "-"? Gordon.

4 5

Re: [Boost-users] Compiling for Solaris, gcc, and STLport
by Tim Clemons 15 Sep '04

15 Sep '04

I found that the following lines in my BOOST_USER_CONFIG allowed for wide character support: #define BOOST_COMPILER_CONFIG "boost/config/compiler/gcc.hpp" #define BOOST_STDLIB_CONFIG "boost/config/stdlib/stlport.hpp" #define BOOST_NO_STDC_NAMESPACE #define BOOST_PLATFORM_CONFIG "boost/config/platform/solaris.hpp" #define BOOST_NO_COMPILER_CONFIG #define BOOST_STRICT_CONFIG I'm not sure why I defined BOOST_NO_COMPILER_CONFIG, or if it was even necessary (it was getting pretty late into the night). I think defining BOOST_STRICT_CONFIG was what did it. On Wed, 2004-09-15 at 04:59, John Maddock <john(a)johnmaddock.co.uk> wrote: > I guess this is a configuration error: to see what Boost config macros are > getting defined, cd into libs/config/test then: > > bjam config_info > > you'll then need to dig down into the bin directory until you find > config_info.out which will have the output from the program - it'll tell you > which macros were defined. # > > As for why they were defined - well I guess either > boost/config/stdlib/stlport.hpp or boost/config/stdlib/libstdcpp3.hpp must > be setting them depending on whether you're using STLPort or not. My guess > is that STLPort hasn't had wide character support enabled for that platform > yet, and that the Boost config system is picking up on that and disabling it > for boost as well. > > Let me know what you find, > > John.

1 0

BGL Graph Strong Components with VC 6
by Daren Grant 15 Sep '04

15 Sep '04

Greetings all , I know this is an odd question especially as the current version of the boost graph library is completely unuseable with VC 6 , but I am using boost_1_27_0 atm , and using the depth_first_search without problems , I have the requirement to also use the strong_components( calls but unfortunately all I get is INFERNAL COMPILER ERRORS ... sorry that should have read INTERNAL ;) ..... does anyone know of a version of the BGL where dfs , and strong_components worked on VC 6 , or a url where i can get archived test results from previous boost builds on specific compilers ? Before I get the string of update to vc 7.1 , I'd love to but my hands are tied :( MTIA

1 1

multi_index - type problems
by Lars Kunert 15 Sep '04

15 Sep '04

Hi! for some reason these two typdefs do not work in my code: typedef typename PriorityQueue::nth_index<0>::type DissimilarityIndex; typedef typename PriorityQueue::index<DissimilarityT>::type DissimilarityIndex; When I actually use them: const DissimilarityIndex& di =_priority_queue.get<DissimilarityT>(); I get error-messages like this: ../../../pocket_graph/src/clustering/NearestNeighbors.h:300: error: invalid initialization of reference of type 'typename boost::multi_index::multi_index_container<Value1, IndexSpecifierList1, Allocator1>::nth_index<0>::type&' from expression of type 'const boost::multi_index::detail::ordered_index<boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, float, &NearestNeighborPair<uint>::dissimilarity>, std::less<float>, boost::multi_index::detail::ordered_index<boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, uint, &NearestNeighborPair<uint>::object>, std::less<uint>, boost::multi_index::detail::ordered_index<boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, uint, &NearestNeighborPair<uint>::nearest_neighbor>, std::less<uint>, boost::multi_index::detail::index_base<NearestNeighborPair<uint>, boost::multi_index::indexed_by<boost::multi_index::ordered_non_unique<boost::multi_index::tag<DissimilarityT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, float, &NearestNeighborPair<uint>::dissimilarity>, boost::multi_index::detail::null_arg>, boost::multi_index::ordered_unique<boost::multi_index::tag<ObjectT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, uint, &NearestNeighborPair<uint>::object>, boost::multi_index::detail::null_arg>, boost::multi_index::ordered_non_unique<boost::multi_index::tag<NearestNeighborT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::const_mem_fun<NearestNeighborPair<uint>, uint, &NearestNeighborPair<uint>::nearest_neighbor>, boost::multi_index::detail::null_arg>, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, std::allocator<NearestNeighborPair<uint> > >, boost::multi_index::tag<NearestNeighborT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::detail::ordered_non_unique_tag>, boost::multi_index::tag<ObjectT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::detail::ordered_unique_tag>, boost::multi_index::tag<DissimilarityT, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::multi_index::detail::ordered_non_unique_tag>' The following typedef works: typedef typename boost::multi_index::index<PriorityQueue,DissimilarityT>::type DissimilarityIndex; The first two typedefs are in some way the more "natural" way to use indices and I wonder if I have just missed to set some preprocessor symbol... ...I am using gcc 3.2.2 Thanks, Lars

2 1

bgl: Walking through two adjacency lists in parallel
by Lars Kunert 15 Sep '04

15 Sep '04

Hi! I am using a graph with setS as EdgeList template-parameter. I currently try to implement to get the number of neighbors which two vertices have in common. I tried to solve this problem by walking thought both out_edge_iterator - ranges in parallel. tie( i, i_end ) = out_edges( vertex_i, graph ); tie( j, j_end ) = out_edges( vertex_j, graph ); int a =0; for( ;; ) { if( *i == *j ) { ++i; ++j; ++a; if( i == iend ) return a; if( j == jend ) return a; } else if( *i < *j ) { // PROBLEM ++i; if( i == iend ) return a; } else /* ( *i > *j ) */ { ++j; if( j == jend ) return a; }; }; return 0; }; And get the following error message: ../../../pocket_graph/src/clustering/GraphAssociation.h:147: error: no match for 'operator<' in 'boost::iterator_facade<I, V, TC, R, D>::operator*() const [with Derived = Which basicaly tells me that operator< is not defined for an edge descriptor. After digging through some bgl-code I found the class stored_edge (in detail/adjacency_list.hpp ) template <class Vertex> class stored_edge : public boost::equality_comparable1< stored_edge<Vertex>, boost::less_than_comparable1< stored_edge<Vertex> > > { public: typedef no_property property_type; inline stored_edge() { } inline stored_edge(Vertex target, const no_property& = no_property()) : m_target(target) { } // Need to write this explicitly so stored_edge_property can // invoke Base::operator= (at least, for SGI MIPSPro compiler) inline stored_edge& operator=(const stored_edge& x) { m_target = x.m_target; return *this; } inline Vertex& get_target() const { return m_target; } inline const no_property& get_property() const { return s_prop; } inline bool operator==(const stored_edge& x) const { return m_target == x.get_target(); } inline bool operator<(const stored_edge& x) const { return m_target < x.get_target(); } //protected: need to add hash<> as a friend static no_property s_prop; // Sometimes target not used as key in the set, and in that case // it is ok to change the target. mutable Vertex m_target; }; for some reason I can not use operator<() or get_target(), but accessing m_target directly works (cause its mutable???)... } else if( i->m_target < j->m_target ) { // SOLUTION ?! However - I feel a bit uneasy to use such a low level interface... Does anyone know of a better solution? Thanks for your help! Lars

2 1

Compiling for Solaris, gcc, and STLport
by Tim Clemons 15 Sep '04

15 Sep '04

On attempting to link against the boost_regex libraries, I'm finding that regex methods and template functions that take wchar_t are not resolving. Examining the libraries with nm show that these methods are not present. My understanding is that wchar_t support for Solaris was only recently introduced in version 3.4.x, so I wondering if the bjam files are taking this into account yet. I'm currently specifying the build with the following invocation: # bjam "-sTOOLS=gcc-stlport" "-sBUILD=release debug <cxxflags>-fPIC" Can I add to this to force the build to use wchar_t? Appreciate the help. -- Tim.

2 1

Problems when linking to regex library.
by Guan Bin 14 Sep '04

14 Sep '04

In my program, I used regex library. When I build the program to the release version, everything is ok. But when I build the debug version, some link error occurs: --------------------------------------------------------------- error LNK2001: unresolved external symbol "void * __cdecl boost::re_detail::get_mem_block(void)" (?get_mem_block@re_detail@boost@@YAPAXXZ) error LNK2001: unresolved external symbol "void __cdecl boost::re_detail::put_mem_block(void *)" (?put_mem_block@re_detail@boost@@YAXPAX@Z) --------------------------------------------------------------- the tools I use: VC6 + sp5 boost_1_31_0 the regex build process: nmake -fvc6.mak nmake -fvc6.mak install Why the link error occurs? How can I fix it? --

2 2

installing limited libraries
by Andrew Dennison 14 Sep '04

14 Sep '04

How do you have bjam install a very limited set of libraries (ie one library only) without having to explicitly list all libraries except the one or two that you might want with a --without-<library> flag?

2 1

Boost on Solaris 8
by cswiger 14 Sep '04

14 Sep '04

Hi - I need the boost libraries for gnuradio, which I'm trying to install on Solaris 8. Are there any prebuilt packages, or step-by-step detailed instructions for Sol8? I made a stab at it but got horribly lost, can post gory detail if necessary. Just curious if there is a simple solution first. TIA --Chuck

2 1

Building libraries on Mac with CodeWarrior 9?
by Jane Lin 14 Sep '04

14 Sep '04

Hello, I haven't been able to find an answer in the archives. I had no problems building boost using bjam for the darwin (gcc) toolset, but have had no luck with CodeWarrior. I posted to the jamboost list, but received no response, so I'm hoping this is the right list! I have: CW 9.2 OS X 10.3.5 boost_1_31_0 boost-build boost-jam-3.1.10-1-macosxppc When I call bjam from the command line, both cwpro9 and cw toolsets (cw-tools.jam downloaded from CVS) result in the same errors: all the codewarrior jam files use -warn when the option should be -warnings, and there is no keyword "nounusedexpr". I edited one of the jam files to change -warn to -warnings and commented out nounusedexpr, but then greg_month.o doesn't get built and subsequently none of the libraries: Error : Nothing to do: no source or object files specified Then I tried a suggestion I saw in the archives - created an empty project for a static library in CodeWarrior and added the files in /boost_1_31_0/libs/regex/src/. This worked fine for the regex library (had to add the boost_1_31_0 path to the access paths in the project settings), but I am getting errors when I do the same for the filesystem library: Error : undefined identifier 'ETXTBSY' exception.cpp line 126 { ETXTBSY, fs::busy_error }, Error : undefined identifier 'lstat' operations_posix_windows.cpp line 318 return ::lstat( ph.native_file_string().c_str( Error : undefined identifier 'S_ISLNK' operations_posix_windows.cpp line 319 && S_ISLNK( path_stat.st_mode ); Error : undefined identifier 'ssize_t' operations_posix_windows.cpp line 489 ssize_t sz; Error : undefined identifier 'sz' operations_posix_windows.cpp line 490 while ( (sz = ::read( infile, buf.get(), bu Error : undefined identifier 'sz' operations_posix_windows.cpp line 491 && (sz = ::write( outfile, buf.get(), Error : undefined identifier 'sz' operations_posix_windows.cpp line 496 if ( sz != 0 ) Error : undefined identifier 'pathconf' operations_posix_windows.cpp line 509 long path_max = ::pathconf( ".", _PC_PATH_MAX ); Is there something else I need to include or define? Thanks a lot! Jane

2 2

statically linking to Boost.Threads with OSX (darwin)
by Simmons, Aaron 14 Sep '04

14 Sep '04

I've built and installed boost on a Mac OSX system (with the 'darwin' toolset). I would like to use static linking. However, the installer is only building the .dylib version of libboost_thread and not the .a version. I thought only Win32 was forced to use dynamic linking? Is there a similar issue on OSX? Thanks, Aaron Simmons

3 3

boost::ublas and gaussian elimination
by Mattias Brändström 13 Sep '04

13 Sep '04

Hello! I would like to do gaussian elimination using uBLAS. It seems that there are many of the tools needed to do gaussian elimination present but I can't actually find a solver. Is there one in uBLAS or has anyone here written one using uBLAS? Thanks in advance! Regards, Mattias

2 1

Re: printing out the full path
by Fred J. 13 Sep '04

13 Sep '04

I think I gave up, I tried every thing and read all I can, I am not able to print out the full path for files in a given directroy, here is another try, cann't figure out whats wrong. any idea why? thanks #include "/usr/include/boost/filesystem/operations.hpp" #include "/usr/include/boost/filesystem/fstream.hpp" #include "/usr/include/boost/filesystem/path.hpp" #include <iostream> #include <string> namespace fs = boost::filesystem; int main(){ fs::path full_path( fs::initial_path() ); char* some_path = "/home/sam/"; full_path = fs::system_complete( fs::path( some_path, fs::native ) ); fs::directory_iterator end_itr; for ( fs::directory_iterator itr( full_path ); itr != end_itr; ++itr ) { std::cout << full_path.native_file_string() << std::endl; } } __________________________________ Do you Yahoo!? Yahoo! Mail - 50x more storage than other providers! http://promotions.yahoo.com/new_mail

2 1